Prompt With Jarvis
Work with meGet the free pack
All guides

FRAMESYou commented this keyword

Watch

A Claude Code skill that turns a video link or file into timestamped frames and a transcript, so Claude can summarize or answer questions about it.

Before you start:
  • Claude Code (desktop Code session, VS Code or terminal)
  • Python 3.10+, FFmpeg and yt-dlp
  • Optional: free Google AI Studio key
Setup
10 min
Level
Intermediate
Stars
18.1k
License
MIT
Jump to the install commandView on GitHub

What it does

Claude can't watch video on its own. Watch fixes that: give it a URL or a video file and it pulls out still frames with timestamps plus a transcript, using the video's own captions when it has them.

The fastest option hands the whole video to Google's Gemini with a free API key. Without a key, everything runs on your computer instead.

Install it

  1. Open Claude Code in a terminal (type claude).
  2. Enter the two commands below inside the Claude Code session, one at a time.
  3. If Watch is not available yet, start a new Claude Code session. Type /watch and pick /watch:watch.
  4. Using Claude Desktop? Click Customize, then Plugins, Add, Add marketplace, Add from a repository. Paste https://github.com/bradautomates/claude-video, click Sync, then install Watch and use it in a Code session.

Copy this

Claude Code
/plugin marketplace add bradautomates/claude-video
/plugin install watch@claude-video

Use it

  1. Optional but fastest: get a free key at aistudio.google.com/apikey and give it to the setup wizard when asked. YouTube links then need nothing else installed.
  2. Put a short video in your project folder and paste the prompt below. Change example.mp4 to your file name.
  3. If Claude says a tool is missing, ask it: "Use the watch skill's bundled setup.py to check dependencies in this session. Tell me which tools are missing and help me install them here."

Your first video

Prompt
Use the watch skill on example.mp4 in the folder I shared.
For setup, choose balanced detail and captions only (none).
Summarize what is visible, with timestamps.

Good to know

Watch does not work in Claude Chat or Cowork, only in Claude Code. It needs Python 3.10+, FFmpeg and the latest yt-dlp; on a Mac, brew install python ffmpeg yt-dlp covers it. Windows install commands are in the README.

On a Gemini run, your video is sent to Google and deleted after the answer. Choose the local engine for private videos.

Links

Want the next one first?

One new guide a week, straight to your inbox.

More like this